Copula modeling of gene coexpression in single-cell RNA sequencing data.

Area of Science:

• Computational biology
• Genomics
• Statistical modeling

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for cellular-level biological studies.
• Accurate statistical models are essential for analyzing scRNA-seq data.
• Gene coexpression modeling is underexplored in scRNA-seq compared to individual gene expression.

Purpose of the Study:

• To evaluate the utility of copula models for gene coexpression in scRNA-seq data.
• To compare the accuracy and efficiency of six different copula models.
• To identify optimal copula approaches for scRNA-seq analysis.

Main Methods:

• Evaluation of six copula models on diverse scRNA-seq reference datasets.
• Assessment of model accuracy in reproducing gene coexpression patterns.
• Analysis of computational efficiency (fitting time) for each model.

Main Results:

• Gaussian copulas demonstrated the best balance of accuracy and computational speed.
• More flexible copula models offered marginal accuracy improvements at a significantly higher computational cost.
• Vine copulas showed potential for high accuracy but current implementations do not scale to large scRNA-seq datasets.

Conclusions:

• Gaussian copulas are a practical choice for modeling gene coexpression in scRNA-seq data.
• The trade-off between copula model complexity, accuracy, and computational efficiency must be considered.
• Further development is needed for scalable vine copula implementations for large single-cell datasets.